Getting it into your agent
One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.
npx agentmods add commands/boxlogodev/sapstack/sap-session-next-turngit clone --depth 1 https://github.com/BoxLogoDev/sapstackWhat it costs to keep this loaded
Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.00054 | $0.02294 |
| Opus 5 | $0.00027 | $0.01147 |
| Sonnet 5 | $0.00011 | $0.00459 |
| Haiku 4.5 | $0.00005 | $0.00229 |
Grade A, and why
sap-session-next-turn scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 2d ago.
A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 223 lines — stays where its author put it; the contents beside it link to each section on GitHub.
/sap-session-next-turn — 다음 AI 턴 실행
입력: $ARGUMENTS
Evidence Loop의 엔진입니다. 세션 상태를 읽고 다음에 실행해야 할 정당한 턴 을 자동으로 결정해 수행합니다. 운영자는 "다음 진행"만 요청하면 되고, AI가 Turn 2/4 구분과 출력 포맷 선택을 책임집니다.
🧭 턴 결정 로직
load state.yaml
switch state.status:
case intake:
→ Turn 1 마무리 & Turn 2 시작 (Hypothesis)
case hypothesizing:
→ Turn 2 (Hypothesis) 재실행 불가 — 이미 진행 중
→ 상태 이상이므로 운영자에게 알림
case awaiting_evidence:
→ 새 Bundle이 없으면 거부
→ Bundle이 있으면 Turn 3 자동 처리 후 Turn 4 (Verify)
case verifying:
→ Turn 4 (Verify) 실행
case resolved:
→ 이미 완료 — reopen 권유
case escalated:
→ 인계된 전문가에게 맡길 것을 상기
case abandoned:
→ 재개할 수 없음. 새 세션 권유.
case reopened:
→ Turn 2 재진입 (새 증거 바탕으로)
--force-hypothesize나 --force-verify는 디버그용입니다. 일반 운영에서
사용하지 마세요. 무결성이 깨질 수 있습니다.
🎛 Turn 2 (Hypothesis) 실행 상세
2-1. 증거 종합
state.bundles[]전부 로드- 모든
items[]를 종류별로 그룹화 - 각 아이템의
source,tags,interpretation힌트 수집
2-2. symptom-index 힌트 적용 (있으면)
state.initial_symptom.matched_symptom_index_entry가 있으면 해당 엔트리의first_check_tcodes,typical_causes,localized_checks를 가설 생성의 시드로 사용
2-3. 가설 생성 (2-4개)
plugins/sap-session/skills/sap-session/references/turn-formats.md의
Turn 2 섹션 규칙을 정확히 따름:
- 각 가설에
falsification_evidence최소 2개 (필수) - confidence 분포 다양화 (전부 0.8+ 금지)
- 최고 confidence 0.95 상한
- 최소 1개는 "컨트롤 가설" (재시도로 해결 같은 단순 경로)
consultant_agents_to_involve필수 지정
2-4. 컨설턴트 에이전트 소환 (선택적 — 이 단계에서도 가능)
가설 생성 단계에서 이미 관련 모듈의 컨설턴트를 호출하면 가설 품질이 올라갑니다. 단, 호출은 병렬로, 각 에이전트에게 현재 Bundle을 요약해서 전달.
병렬 호출 파이프라인:
for agent in hypothesis.consultant_agents_to_involve:
spawn Agent(subagent=agent, context={bundles, symptom, country})
await all
소환 결과는 가설의 technical_chain을 풍부하게 하는 데 사용. 이 단계에서
에이전트가 "이 가설은 말이 안 된다"고 반론하면 가설 confidence 하향.
2-5. Follow-up Request 생성
plugins/sap-session/skills/sap-session/references/followup-authoring.md의
5가지 품질 기준(최소성·반증 중심성·비용 정직성·안전성·명확성)을 모두 준수.
2-6. 상태 전이
hypotheses/h-*.yaml에 각 가설 저장requests/flr-*.yaml에 Follow-up Request 저장state.hypotheses[],state.followup_requests[]에 appendturns[]에 Turn 2 엔트리 (complete) + Turn 3 엔트리 (pending)status: intake → awaiting_evidence(Turn 2+3를 한 번에 셋업)audit_trail:hypothesis_proposed,followup_requestedcurrent_turn_number: 3
What this file has done since we first saw it
Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.
- 2d ago First seen · 223 lines · 54 tokens per session scan A 48aed5f244e1
sap-session-next-turn is a command published in the GitHub repository BoxLogoDev/sapstack (18 stars, last pushed 3d ago), licensed MIT. It adds 54 tokens to every session and 2,294 once invoked, about $0.0003 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other commands, from other repositories
eval-author
Draft a new eval case (eval/cases/ .json) + scaffold its golden folder, goldenpending until captured on the VM.
ingest-l1
L1 analysis loop for the abapwiki knowledge base: for each batch it launches the abap-analyzer sub-agent in parallel, then the adversarial judge abap-deepcheck (separate session), applies only the analyses that pass the fail-closed gate, and commits. Resumes exactly after an interruption. Use this skill to document…
btp-architecture-review
Review SAP BTP architecture decisions for account model, services, security, integration, operations, and rollout readiness.
work-zone-content-check
Check SAP Build Work Zone Advanced content packages, site artifacts, roles, destinations, and transport readiness.
btp-destination-diagnose
Diagnose SAP BTP destination and Cloud Connector issues using configuration checks and safe troubleshooting steps.
sap-kickoff
Trigger project kickoff for a new SAP engagement — generates a full project charter with scope, governance structure, RACI matrix, and risk register.